THPPH070 Proceedings of FEL 2006, BESSY, Berlin, Germany

OPTIMUM BEAM CREATION IN PHOTOINJECTORS USING

SPACE-CHARGE EXPANSION∗

J.B. Rosenzweig, A.M. Cook, M. Dunning, R.J. England

Particle Beam Physics Laboratory, Department of Physics and Astronomy
University of California Los Angeles, Los Angeles, CA 90095, USA
M. Bellaveglia, M. Boscolo, L. Catani, A. Cianchi, G. Di Pirro,
M. Ferrario, D. Filippetto, G. Gatti, P. Musumeci, L. Palumbo, C. Vicario
Istituto Nazionale di Fisica Nucleare, Laboratori Nazionale di Frascati
00044 Frascati (Roma), Italy
S. Jones
Jet Propulsion Laboratory, California Institue of Technology
4800 Oak Grove Drive, Pasadena, CA 91101, USA.

Abstract
It has recently been shown that by illuminating a photo-
cathode with an ultra-short laser pulse of appropriate trans-
verse profile, a uniform density, ellipsoidally shaped bunch
is dynamically formed, which then has linear space-charge
fields in all dimensions inside of the bunch. We study here
this process, and its marriage to the standard emittance
compensation scenario that is implemented in most modern
photoinjectors. It is seen that the two processes are compat-
ible, with simulations indicating that a very high brightness
beam can be obtained. The scheme has produced stimulus
for a series of experiments at the SPARC injector at Fras-
cati in 2006-2007. An initial time-resolved experiment has
been performed involving Cerenkov radiation produced at
an aerogel. We discuss the results of this preliminary ex-
periment, as well as plans for future experiments to resolve
the ellipsoidal bunch shape at low energy. Future measure-
ments at high energy based on fs resolution RF sweepers
are discussed.
INTRODUCTION
In order to obtain the highest brightness electron beams
from photoinjectors, it is most common to rely on the emit-
tance compensation process [1]. Optimization of this pro-
cess demands that the transverse fields be as uniform, and
linear (in radius r) as possible. The existing studies of emit-
tance compensation have, to that end, assumed use of a
uniform density electron beam, having a cylindrical shape.
However, this shape produces space-charge fields near the
beam head and tail that have pronounced nonlinear depen-
dencies on the spatial coordinates. These nonlinearities re-
sult in both transverse and longitudinal emittance growth.
It has been known for some time, however, that a uni-
form ellipsoidal density distribution yields space-charge
∗ Work supported by the US Dept. of Energy under grant DE-FG03-
92ER40693, and the Italian Ministry.
fields that are linear in all dimensions [2]. Under such con-
ditions, it is conceivable that one may obtain essentially
emittance growth-free dynamics. How to produce such a
distribution has, until recently, remained an unanswered
question.
In 1997, Serafini proposed the dynamic creation of an el-
lipsoidal bunch by launching an ultra-short, radially shaped
beam, which then evolves to achieve the desired longitudi-
nal shape [3]. On the other had, it has recently been shown
by Luiten, et al., that in obtaining the correct final ellip-
soidal distribution, there is essentially no requirement on
the shape of the initial laser pulse other than it be ultra-
short (length τl much shorter than eventual beam length
after space charge expansion) [4]. Thus such laser pulses
are a natural, and technically achievable way of producing
an ellipsoidally shaped, nearly uniform density beam.
As the beam dynamics just after photoemission are qual-
itatively different in the traditional emittance compensa-
tion scenario and in the Luiten-Serafini scheme, it is not
immediately apparent that one may successfully combine
the two. The UCLA-SPARC collaboration has recently
shown [5] that this marriage is indeed possible; further, the
combination of emittance compensation and dynamic cre-
ation of the ellipsoidal shaped beam produces results that
in many ways are superior to those obtained in state-of-the-
art designs. As the bunches that are produced are shorter
than in standard cases, very high brightness beam creation
is possible.
The basic idea behind the Luiten-Serafini scheme is sim-
ple: the beam profile expands and deforms longitudinally
to produce, in the final state, a uniformly filled ellipsoid of
charge. In the process, phase space rearrangements occur
which degrade the emittances– especially in the longitu-
dinal dimension. In order to understand this process, to
specify experimental requirements, and to identify experi-
mental signatures associated with the process, we have an-
alyzed the dynamics of space-charge-dominated beam ex-

752 FEL Technology

 Proceedings of FEL 2006, BESSY, Berlin, Germany
pansion [5].
This analysis may be summarized in a few points:
First, the injected bunch surface charge density
σb = dQb /dA must not be too high, or image charge ef-
fects at the cathode distort the final pulse profile so that
it is not ellipsoidal. This is quantified by the condition
α ≡ 4πσb /E0 << 1.
Second, the beam must be much shorter than its eventual
size in order to be able to ignore the details of the initial
pulse profile. In practice, 300 fs laser pulses (typical of
the limitations of the SPARC photocathode drive laser af-
ter conversion to UV) excite roughly the same length elec-
tron bunch, which expands to around 4 psec in our exam-
ple cases. The pulse length after expansion is estimated as
Lb ≈ 2πσb me c2 /E02 .
The current density that is achieved after expansion
is Jz = eE02 /4πme c, a constant dependent only on the
applied electric field E0 . All beams become uni-
form in density. To achieve the desired ellipsoidal
beam shape, one must choose the initial surface cur-
rent density distribution
 correctly, which implies that
σb (r) = (3Qb /2πa2 ) 1 − (r/a)2 .
While the analysis of the beam dynamics is useful, the
central issue of joining this regime with emittance compen-
sation must be explored with simulations. The initial simu-
lations begun in Ref. [5] that we have performed are in the
context of the SPARC scenario, so that we may proceed
directly to discussing the experimental tests of this new
regime– now commonly known as the “blowout regime”–
of the photoinjector there.
BLOWOUT REGIME WITH EMITTANCE
COMPENSATION: GENERAL STUDY
We have performed initial UCLA PARMELA [6] simu-
lations to explore joining the Serafini-Luiten scheme with
the optimized emittance compensation working point of the
SPARC injector at LNF. We assume that the gun (1.6 cell,
2856 MHz) and solenoid are the same, and run near to the
standard conditions. Through trials, we have optimized the
launch conditions of the beam. In order to have values of
α which do not give excessive image charge effects, the
beam charge is lowered and the beam radius is slightly
enlarged. In the preliminary optimization, we launch a
0.33 nC beam with an initial longitudinal Gaussian distri-
bution having σ t = 33 fs beam, and a radial Gaussian with
σx = 0.77 mm (cutoff at 1.8σ). The gun is run with a peak
on-axis gradient of 120 MV/m; the beam is launched at
33 degrees forward of crest. This is a bit advanced in com-
parison to the nominal launch phase for a standard bunch,
and serves to control the excessive beam energy spread af-
ter the gun. The emittance compensation solenoid is run
with peak field Bz = 2700 G, which is slightly below the
standard scenario, as the beam has slightly lower energy
exiting the gun. We note that the peak value of α in our
case is 0.11, as opposed to 0.42 in the LCLS design.
The formation of the quasi-ellipsoidal bunch is clearly
FEL Technology
THPPH070
Figure 1: (left) PARMELA simulation results of electron
bunch (x, z) distribution 133 cm from cathode (6.3 MeV
energy), showing ellipsoidal beam boundary. (right) Evo-
lution of σδp/p in z for emittance compensation case, from
PARMELA simulation.
shown in Fig. 1, which displays the bunch (x, z) distribu-
tion at a point 133 cm from the cathode, in the drift space
after the gun and just preceding initial traveling wave linac
section. Here the beam has 6.3 MeV mean energy, and its
transverse dynamics are space charge-dominated. Thus one
sees clearly the “inflated” ellipsoidal beam shape. The final
bunch length is 1.3 mm full width, corresponding to a peak
current of 105 A. Thus even with one-third of the charge,
this scheme should produce a higher current than obtained
in simulations of the standard design.
Two notable defects are seen in the beam shape in Fig. 1.
The first is the extension of the half-ellipsoid in the trail-
ing part of the bunch as compared with the initial half– an
asymmetry caused by image charge effects. This non-ideal
behavior in fact gives the limit on σ; when one attempts
to launch a higher surface charge density, the bunch defor-
mation from the desired symmetric ellipsoid produces poor
emittance performance. The second notable feature is the
existence of an anomalous ring at the outer radial edge of
the beam. This part of the beam has low surface charge
density and experiences radially fringing fields due to its
edge location. Because of these effects, it does not experi-
ence enough longitudinal expansion to keep pace with the
rest of the bunch, but instead has a moderate amount of
radial expansion.
As the longitudinal space-charge during much of the
acceleration is also linear, and total pulse length T is
short, the longitudinal phase space is very compact. The
evolution of the relative momentum spread σ δp/p in z
is shown in Fig. 1. The final achieved RMS value
is σδp/p = 1.6 · 10−4 , which is an order of magnitude
smaller than that obtained in the standard LCLS type (or
SPARC type) design.
The evolution of the RMS transverse beam size σ x , and
the RMS normalized emittance  n,x are shown in Fig. 2.
While the behavior of σ x is similar in most respects to the
standard design, with the beam approximately matched at
linac entrance to the invariant envelope, the emittance be-
havior is not as familiar. In the standard LCLS design,  n,x
achieves a minimum value in the post-gun drift, rising to a
local maximum at injection into the linac. The focusing and
adiabatic damping of the motion in the linac then produce a
753
THPPH070 Proceedings of FEL 2006, BESSY, Berlin, Germany
Figure 2: (left) The evolution of RMS transverse beam size
σx for emittance compensation case, from PARMELA sim-
ulation. (right) Evolution of RMS normalized emittance
n,x for emittance compensation case.
monotonic decrease of  n,x in z. In our case, the transverse
space-charge and thus the plasma/emittance oscillations do
not “turn on” until after the longitudinal expansion is well
underway, thus delaying the emittance minimum in Fig. 2
to occur inside of the linac. In order to produce faster emit-
tance oscillations in the linac to strongly diminish  n,x be-
fore acceleration removes the plasma-dominated beam be-
havior, the solenoid field in the first linac section has been
raised by 40% relative to the standard scenario. This ploy
works well, as the final value (still slightly decreasing) of
n,x at the end of the second linac (84.5 MeV energy) is
0.68 mm-mrad. The thermal emittance at the cathode is
0.4 mm-mrad, and so the space-charge induced emittance
is well compensated.
After acceleration to higher energy (84.5 MeV), the
beam is not spacecharge dominated, and the (x, z) profile
no longer ellipsoidal. Nonetheless, the beam has excel-
lent emittance,
 and maintains a current profile with shape
I(t) ∝ 1 − (2t/T )2 . With a high initial current, and low
intrinsic energy spread, this beam may be compressed fur-
ther, with very high final peak current achievable [7].
CONSIDERATIONS FOR SPARC
EXPERIMENT
Laser and photocathode issues
Before discussing the planned electron beam measure-
ments, we first review some experimental considerations
specific to the use of such short lasers. First, we note that
the laser pulse should be tailored, either by collimating
and relay imaging, and that it must be quite short. In first
measurements at SPARC, we have found, through cross-
correlation measurements, that the UV pulse is difficult to
make shorter than 300 fs using 3rd harmonic generation
with noncollinear mixing in the conversion crystals. In or-
der to obtain even this result, we must give up UV energy,
going from 1.5 mJ to 0.2 mJ. While this energy is still ad-
equate for obtaining 0.33 nC of charge using a metal cath-
ode (Cu or Mg), to go shorter may be difficult. In order to
check the effect of this extra pulse length on the scheme, we
have performed simulations analogous to the original ex-
ploratory, indications are that the added length in the laser
754
pulse does not affect the final electron beam configuration
at this level.
The laser intensity needed for this scenario is a factor
of 30-100 higher than in the LCLS case. The issue of
laser damage at the cathode surface has been examined,
and found based on previous experience at UCLA and at
the BNL ATF to not be worrisome [7].
Experimental signatures and measurements
In the SPARC experiment, we plan first to image the
beam (time-integrated) at low energy (5-7 MeV) in the re-
gion after the gun, using a YAG detector. For time resolved
measurements we will first convert the beam spatial infor-
mation to photons with a Cerenkov convertor. In order to
have a manageably small angle of emission we use aerogel,
which has a small index of refraction (n = 1.005-1.02).
At 5 MeV the Cerenkov emission threshold is reached for
n = 1.005; we may choose angles of emission from near
zero at this threshold to up to 9 degrees with the aerogels
that are presently in hand. The aerogels have been custom
fabricated at the Jet Propulsion Laboratory.
Simulations consist of providing electrons (typically
40,000) from PARMELA to GEANT [8], which simulates
the scattering of the electrons in the entrance foil and gen-
erates a collection of Cerenkov photons in the aerogel. The
photon distributions that result are then passed to a Mathe-
matica based, optical ray-tracing program, Rayica.
The streak camera at SPARC has 2 ps FWHM time res-
olution, which is on the border of resolution for the bunch
length of interest. This will be handled in initial measure-
ments by use of higher charges (above 1 nC), and thus
longer beams, to test the longitudinal expansion dynamics
of the beam. We thus must consider alternative schemes
based on ultra-fast gating [7].
FIRST RESULTS
The first stage of experimentation on the blowout regime
took place at INFN-LNF beginning at the end of March
2006. During commissioning of the SPARC RF photocath-
ode gun, the UCLA-produced gun was conditioned quickly
up to 11 MW, which produces 110 MV/m peak electric
field, and a 5.7 MeV electron beam.
The laser was reconfigured for short pulses (less than
0.5 psec FWHM) and up to 1.6 nC of charge. The con-
ditions for observing the dynamic creation of nearly uni-
formly filled ellipsoidal charge distributions were not quite
present; in fact, the emittance was not of equivalent quality
to that obtained in standard operation. Nevertheless, im-
pressive first data was obtained.
Initial measurements of the longitudinal-transverse pro-
file of the beam were made with aerogel with index
n = 1.008, with the Cerenkov radiator placed 2.4 m away
from the cathode, downstream of the slit-based emittance
measurement system. Streak camera images were obtained
using the transport system described in the previous sec-
tion. Such a streak, after correction, is shown in Fig. 3.
FEL Technology
 Proceedings of FEL 2006, BESSY, Berlin, Germany THPPH070

Figure 3: Streak image from SPARC experiment, after cor-

rection.

This image displays the profile obtained from a bunch

with charge of 700 pC. A large charge is preferred in this
case in order to discern information at a time scale longer
over the streak camera resolution; the 700 pC case is ex-
pected to have expansion of approximately 7 psec FWHM,
well in excess of this 2 psec FHWM resolution.
Streak images obtained in the highest temporal resolu-
tion mode are inherently noisy; this condition is required Figure 4: Analysis of streak data, with fraction of inte-
in order to avoid space-charge induced pulse distortion in- grated intensity of data inside of elliptical contour shown.
side of the streak tube. Thus in order to extract informa- Best fit of data points to three models are shown: bi-
tion from single shots concerning the streak image– which gaussian distribution, uniform elliptical distribution, and
should represent the beam density distribution in an (x, z) Fermi-Dirac (uniform with tails) distribution.
slice in the midplane of the bunch– we have adopted a max-
imum likelihood analysis to test for different assumed types
70 degree) injection. Additional improvements should also
of beam distributions.
result from use of laser cleaning of the cathode. The
The (x, z) slice distributions we have tested for consis-
SPARC injector will soon be completed with the addition
tency with the data include: (1) a bi-Gaussian (thermal-
of post-acceleration linacs and beam diagnostics (e.g. RF
type) distribution; (2) a uniformly filled ellipse (assumed
sweeper). In this fully mature experimental scheme, a com-
arising from a parent uniformly filled ellipsoid); and (3) a
plete test of the consistency of the Luiten-Serafini scheme
nearly uniformly filled ellipse with a tail, which we choose
with emittance compensation should be possible, using the
to represent as a Fermi-Dirac distribution.
large array of techniques described here. Further experi-
As all of the distributions assumed have contours of con-
ments will also emphasize the demonstration of high qual-
stant density that are elliptical, a systematic statistical ap-
ity longitudinal phase space and concomitant low energy
proach is possible, in which we look at the total integrated
spread, as well as high compressibility.
intensity inside of ellipses of size varying from zero area
to an area covering the entire streak image. These ellipses
are all required to have the same aspect ratio, which is given ACKNOWLEDGEMENTS
by the intensity profile itself, R = σ x /vs σt , where vs is the
The authors acknowledge useful interactions with Scott
streak velocity, and σ t =3.45 psec for the streak in Fig. 3.
Anderson, Luca Giannessi, Marcello Quattromini, and
With these functions in hand, we can fit to the data given
Dirk Lipka.
in the streak images to determine the likelihood that one
of the assumed three profiles is more likely than the others.
Such an exercise has been performed for the streak given in REFERENCES
Fig. 3, with the results shown in Fig. 4. It can be seen the [1] L. Serafini and J.B. Rosenzweig, Physical Review E 55, 7565
bi-Gaussian hypothesis can be rejected as the least likely (1997).
model. While the uniformly filled ellipsoid gives a good [2] I.M. Kapchinskii and V.V. Vladimirskii, in Proc. Int. Conf.
fit near the distribution center, it is not very accurate at the High En. Accs., CERN, Geneva, 274, 1959.
edge, where one expects strong deviations in any case from [3] L. Serafini, AIP Conf. Proc. 413, 321 (1997).
this model. Finally, we note that the best fit obtained from [4] O. J. Luiten, et al., Phys. Rev. Lett., 93, 094802-1 (2004).
the Fermi-Dirac model gives an excellent match to the data.
[5] J. B. Rosenzweig, et al., Nucl. Inst. and Meth. A, 57 87
(2006).
CONCLUSIONS [6] E. Colby, UCLA PhD Thesis, FERMILAB-THESIS-1997-03
(FNAL, 1997).
While the first measurements have established the
[7] J.B. Rosenzweig, et al., Proc. 2005 Phys. App. High Br.
soundness of the basic experimental approach, much more Beams, Erice, 2006.
remains to be done. As of now, the SPARC injector
[8] S. Agostinelli, et al., Nuc. Inst. and Meth. A 506, 250 (2003).
is being modified to allow for near-axis (as opposed to

FEL Technology 755